The AMD Execution Thread [2007 - 2017]

[DavidGraham]

It was certainly not because the Fiji/Hawaii/Tonga cards kept showing better performance than their Maxwell 2 equivalents with practically every new DX12 benchmark that came up. Future-proofing is a totally overblown concept.

So average Joes didn't hear about imminent launch of new GPUs, but heard about a handful DX12 games that favor certain arc and ignored the dozens others that don't favor that arc?

It's okay if you don't care. What's not okay is to use these botched survey results to somehow try to discredit actual GPU sales numbers.

Which has nothing to do with sales, further analysis reveals the claimed market growth is due to overall reduction in GPU sales, not because of a sudden momentum on AMD's part. In fact AMD shipments decreased 1% since last quarter (according to JPR), NV decreased 20% due to clearing old stocks, and Pascal supply issues, but made up for it with very healthy margins.

People seem to be hung on the premise that AMD's hardware percentage should increase in Steam's survey, when In fact it's showing what the current market is at the moment, for the past quarters NV outsold AMD 4:1, new NV GPUs have 400% more probability to appear in the survey. That increase didn't come from nowhere, people shifted their allegiance from AMD to NV, that's why the percentage of certain AMD families eroded away. And it will continue to do so, as owners of enthusiast GPUs switch sides, at least till AMD gives the appropriate response.

Does it really matter anyway? Are we really complaining that "AMD sold more, but not that much more"?

AMD didn't sell more, Q2 is the weakest so numbers show they didn't sell more, their share grew relative to the competition because the competition had supply issues, If by twisting logic this is considered a win for AMD, then by all means, but empty barrels will only get you so far.

 

[Razor1]

Mercury could be less precise than AMD's own accounting and absolute numbers of 2016Q2 are probably higher than 2015Q2.

Does it really matter anyway? Are we really complaining that "AMD sold more, but not that much more"?

AMD cited Mercury and JPR numbers......

They didn't sell more, that is what those numbers show, and Mercury even states AMD's total shipments went down from last quarter if you look at the full report, JPR gave a precise amount of .1% drop for AIB shipments as well for AMD in their public release of their report.

Maybe she misspoke, shipments vs. marketshare? You think something like that is overlooked though?

 

[firstminion]

I have only cursory interest on this so I'm perfectly fine with waiting a couple of quarters to see how this works out.

By what was quoted I'm fairly sure Dr Lisa Su said the truth (or really close to it) and AMD had the biggest shipments in their recent history.

@DavidGraham if Q2 is the weakest and AMD isn't competitive then shouldn't AMD be the one to take the biggest hit?

@Razor1 You made me curious so, is this the full report? http://jonpeddie.com/publications/market_watch

Overall GPU shipments (rounded up) decreased -4% from last quarter, AMD decreased -1% Nvidia decreased -20% and Intel, decreased -1%

Last time I saw Intel didn't had any Desktop Discrete GPU so... not comparable. BTW:

AMD's shipments of desktop heterogeneous GPU/CPUs, i.e., APUs, for desktops decreased -22.2% from the previous quarter. AMD's shipments were down -0.4% in notebooks. Desktop discrete GPUs increased 4.03% from last quarter, and notebook discrete shipments decreased -3.7%. AMD's total PC graphics shipments decreased -1.4% from the previous quarter.

AMD increased shipments from a stronger to the weakest quarter... It's right there on the report

 

[Razor1]

Firstminion, the problem with this now is you are using JPR numbers, JPR numbers are not what AMD internal numbers should be, Mercury numbers are, this is why I've stated they mixed and matched numbers across two totally different ways of measuring marketshare.

Its all fake lol. You can't take JPR numbers for one one thing and then take Mercury numbers on another, and compare the two to get what you want!

Either you talk about AMD shipments (Mecrury) or you talk about AIB shipments (JPR), not both, and come out with what they stated.

I was not only confused what Lisa Su stated because I was looking at the numbers shortly after, I couldn't even make the numbers fit with what was being stated,even with the PR spiel they had was worse.

The last quarter I have been thinking about investing in AMD stock, but I can't do that with things like this happening.

When I invest, I don't care about what the CEO says, or other board members, or analysts say, everyone in those conference calls, reporters, all have agendas, so I do my own research and compare it to what they are saying and their financials.

Its a risky stock as it is without this, even though the money I would put into AMD stock is "at my own risk" and am willing to possible loose if it happens, but I don't try to put too much risk into it.

 

[madyasiwi]

The shipment number discrepancies might have something to do with Apple.

 

[Grall]

The shipment number discrepancies might have something to do with Apple.

Yeah, Apple is allegedly updating their Mac lineup, and might even be the reason RX 4xx cards are so hard to find in the channel. They're probably sucking up a large, perhaps the major portion of the chip supply to feed their war machine assembly lines...

 

[Razor1]

Yeah, Apple is allegedly updating their Mac lineup, and might even be the reason RX 4xx cards are so hard to find in the channel. They're probably sucking up a large, perhaps the major portion of the chip supply to feed their war machine assembly lines...

nope, apple doesn't do many discrete cards at all, actually they don't even procure the cards anymore... the Apple outlets do that now. Also its a very small amount, what Apple buyers needs graphics cards? graphics guys and motion picture stuff, that's it, anyone else buying an apple just doesn't need one and the way apple prices those types of addons out (way over priced).....

Also Apple doesn't buy from AIB's, those numbers shouldn't tally into AIB numbers.

All Mac Pro's are custom fit boards, special ordered from foxconn, all I-macs are using custom mxm modules mobile versions of rx 4xx series which we know are not rx4xx chips, but last gens chips.

 

[Grall]

nope, apple doesn't do many discrete cards at all, actually they don't even procure the cards anymore...

I didn't say Apple does discrete cards. My point, more detailed: assuming Apple is buying up a large portion of the available chip supply (and Apple does use discrete GPUs in iMacs, Macbook Pros and now potentially also the rumored new thunderbolt display with built-in discrete graphics), well, then you can't make discrete cards out of these same chips can you?

 

[Razor1]

Q2 apple shipment numbers have dropped, hard to say its because of them too. I suspect they don't do more than low hundred thousand units for add in boards in a quarter, maybe even less, this is why we never see marketshare fluctuations when apple changes vendors in the graphics card markets.

Outside of the macpro, I don't think they use the rx4xx, and by far it is the lowest selling systems they have.

Even design studios that used to use Macpro towers have switched over to imacs. The macpro cylinder is not picked up because they are harder to work with for the it departments (not harder but in the sense of upgrading, repair, etc).

PS I can't even find the rx480 in those either they are selling with FirePro D700 which are last gen cards too.

Here are the configurations for the updated Imac

Included Software
Mac OS X El Capitan (Includes Pages, Numbers, Keynote, iPhoto, iMovie, GarageBand, Mail, Address Book, iCal, the Mac App Store, iTunes, Safari, Time Machine, FaceTime, Photo Booth, Launchpad, AirDrop, Resume, Auto Save, Versions, QuickTime, Mail, Address Book, iCal, and more)

Configuration Options
Memory up to 32GB 1600MHz DDR3 SDRAM
Hard Drive up to 3TB Fusion Drive or 1TB of flash storage
Intel HD Graphics 6000
Intel Iris Pro Graphics 6200
AMD Radeon R9 M380 with 2GB of GDDR5 memory
AMD Radeon R9 M390 with 2GB of GDDR5 memory
AMD Radeon R9 M395 with 2GB of GDDR5 memory
AMD Radeon R9 M395X with 4GB of GDDR5 memory
Windows and Virtualization
Service Plans (Apple Care, LoJack, Service Net)
Software and Accessories

here are the configeration options for the updated mac pro

Configuration Options:
• Configurable to 3.5GHz 6-core processor with
12MB L3 cache, 3.0GHz 8-core processor with
25MB L3 cache, or 2.7GHz 12-core processor with
30MB L3 cache
• Configurable to 16GB (four 4GB), 32GB (four
8GB), or 64GB (four 16GB)
• Configurable to dual AMD FirePro D500 with 3GB
of GDDR5 VRAM each or dual AMD FirePro D700
with 6GB of GDDR5 VRAM each
• Configurable to 512GB or 1TB PCIe-based flash
storage

Those are the only two that use discrete cards.

As of right now they don't even have drivers for the rx480 in mac's, and not expected to have drivers till end of this year.

 

[Grall]

Outside of the macpro, I don't think they use the rx4xx

I thought there was a news post about Apple selecting the rx 4x0 (don't remember which one) a while back. *shrug*

 

[Razor1]

well haven't seen it available in their systems yet, its already back to school sales too.

 

[Pressure]

nope, apple doesn't do many discrete cards at all, actually they don't even procure the cards anymore... the Apple outlets do that now. Also its a very small amount, what Apple buyers needs graphics cards? graphics guys and motion picture stuff, that's it, anyone else buying an apple just doesn't need one and the way apple prices those types of addons out (way over priced).....

Also Apple doesn't buy from AIB's, those numbers shouldn't tally into AIB numbers.

All Mac Pro's are custom fit boards, special ordered from foxconn, all I-macs are using custom mxm modules mobile versions of rx 4xx series which we know are not rx4xx chips, but last gens chips.

Just to clear something up - Apple does not use MXM modules in their iMac line-up but solder the GPU right on the logic board.

 

[Razor1]

when did they change that?

Was working on a imac last year (parents) it had an mxm, then again, it could have been a year old at that time.
Must have happened fairly recently till 2014 there are still parts for MXM III cards for AMD Imacs.

 

[Pressure]

when did they change that?

Was working on a imac last year (parents) it had an mxm, then again, it could have been a year old at that time.
Must have happened fairly recently till 2014 there are still parts for MXM III cards for AMD Imacs.

The last iMac with an MXM slot were from 2011. 2012 onwards were soldered directly to the logic board.

 

[firstminion]

@Razor1, my bad. Mercury, then:

Overall unit shipments of PC graphics fell by 6.5 percent in the second quarter of 2016, and were down 3.1 percent compared to a year ago. Due to a strong shift towards high-end graphics driven both by new products launched in the quarter and fewer low-end units shipped, average desktop selling prices surged and set a new record high, resulting in a substantial increase in GPU revenues for the quarter even as the units declined. With the exception of chip sets, which experienced an increase due to end-of-life sales, all major graphics segments in both desktops and notebook PCs declined in unit terms. The strongest declines were in discrete desktop GPUs, shipments of which dropped more than 20 percent compared to the prior quarter, due to inventory correction. Only Nvidia experienced declines in this segment as AMD had modest growth. Paradoxically, discrete desktop GPUs were up onyear for the second quarter in a row due to weak shipments in the year-ago quarter.

AMD’s average total share increased for the first time in more than four years. The last time the company experienced an increase was the first quarter of 2012. Both desktop and mobile shares were up this quarter.

http://www.mercuryresearch.com/graphics-pr-2016-q3.pdf

 

[Razor1]

@Razor1, my bad. Mercury, then:





http://www.mercuryresearch.com/graphics-pr-2016-q3.pdf

A year ago How about in Q1 of 2015? It was higher then, Q3 of 2014 it was higher then, Q4 of 2014 it was higher then too!

That is the problem, the CEO stated it was the highest shipments they have had since 4th q 2014. I find that hard to believe when prior to the major marketshare lose in end of 2014, AMD was doing fine. It was only after then Maxwell 2 line started, Sept of 2014 did AMD start loosing marketshare. But the huge slide didn't come up till end Q1 of 2015 which would not shown up till the Q2 numbers! So what are they saying, they didn't tell the truth back when Maxwell 2 was out for quarter?

BTW you just quoted the mercury report I quoted last page, actually the same paragraph.

Look at Mercury numbers for the past year and half, you will see that trend. Actually I think I posted it already.

yes, same post that quote you quoted from mercury report where I quoted it too.

Now prior to Q2 2015, Q1 showed that AMD had even more marketshare with a NON WEAK quarter from a total shipment point of view

Now if you mix in JPR numbers, into the mercury numbers, I can see how to make up what she stated as AIB shipments shifted differently from IHV shipments during the quarterly slides. Which is to be expected because there needs to be cause and effect for AIB's to change the amount of chips they order.

What AMD is saying is in a historical low quarter for AIB shipments, they out shipped stronger quarters with less marketshare % overall? Does that make sense? I find that extremely hard to believe given the type of market it is (mature and saturated, in such a market things like that don't happen unless there is something like supply chain issues that are not fore seen).

Edit:

This is not in the margin of error either, cause we are talking about 2% shift in total marketshare with a 6.5% drop in total volume (from the weakest Quarter last year which was Q2 2015. We are talking probably in the neighborhood of 500k units. The only way AMD's statements are even remotely right, is if there was a large drop in IGP and notebook discrete and a large uptake of desktop discrete, and we know that didn't happen. The math just doesn't work out unless there are errors with what they have stated, or the numbers are just wrong. I think that would be clarified with Q3 numbers.

 

[firstminion]

A year ago How about in Q1 of 2015? It was higher then, Q3 of 2014 it was higher then, Q4 of 2014 it was higher then too!

That is the problem, the CEO stated it was the highest shipments they have had since 4th q 2014. I find that hard to believe when prior to the major marketshare lose in end of 2014, AMD was doing fine. It was only after then Maxwell 2 line started, Sept of 2014 did AMD start loosing marketshare. But the huge slide didn't come up till end Q1 of 2015 which would not shown up till the Q2 numbers! So what are they saying, they didn't tell the truth back when Maxwell 2 was out for quarter?

BTW you just quoted the mercury report I quoted last page, actually the same paragraph.

Look at Mercury numbers for the past year and half, you will see that trend. Actually I think I posted it already.

yes, same post that quote you quoted from mercury report where I quoted it too.

Now prior to Q2 2015, Q1 showed that AMD had even more marketshare with a NON WEAK quarter from a total shipment point of view

Now if you mix in JPR numbers, into the mercury numbers, I can see how to make up what she stated as AIB shipments shifted differently from IHV shipments during the quarterly slides. Which is to be expected because there needs to be cause and effect for AIB's to change the amount of chips they order.

What AMD is saying is in a historical low quarter for AIB shipments, they out shipped stronger quarters with less marketshare % overall? Does that make sense? I find that extremely hard to believe given the type of market it is (mature and saturated, in such a market things like that don't happen unless there is something like supply chain issues that are not fore seen).

Edit:

This is not in the margin of error either, cause we are talking about 2% shift in total marketshare with a 6.5% drop in total volume (from the weakest Quarter last year which was Q2 2015. We are talking probably in the neighborhood of 500k units. The only way AMD's statements are even remotely right, is if there was a large drop in IGP and notebook discrete and a large uptake of desktop discrete, and we know that didn't happen. The math just doesn't work out unless there are errors with what they have stated, or the numbers are just wrong. I think that would be clarified with Q3 numbers.

The report says, right there, that "Paradoxically, discrete desktop GPUs were up onyear for the second quarter in a row due to weak shipments in the year-ago quarter."

Don't misuse percentages, if you want to prove something you should work harder than "I find that extremely hard to believe". Get actual absolute figures.

Anyway I'll quote myself: "I have only cursory interest on this so I'm perfectly fine with waiting a couple of quarters to see how this works out."

 

[Razor1]

here are the share values for 2015 q1 and back.

http://www.mercuryresearch.com/graphics-pr-2015-q2.pdf

AMD is talking out of their ass.

These numbers are arbitrary numbers.

If total shipped of GPU were 10 million, and ya had 28% market share then you have 2.80 million units right?

If now you have 12 million total shipped and 23% market share 2.760 million units?

Get the picture, this isn't margin of error.

Q2 of last quarter was worse than Q2 of this quarter.

The Paradoxically statement, is as a whole discrete market. And from the info you have that you are privy to Q1 of 2015 it wasn't weak not as weak as this recent quarter.

 

[Silent_Buddha]

Why are you comparing Q1 2015 values with Q2 2016 values? You either compare Q1 2016 to Q2 2016 (quarter to quarter) or you compare Q2 2015 to Q2 2016 (year over year).

This is the one you should be using.

http://www.mercuryresearch.com/graphics-pr-2015-q3.pdf

• Q2 of 2015 has overall shipments down 21.0% compared to the previous year. Nvidia saw growth in desktop discrete shipments while AMD declined massively in desktop discrete shipments. Desktop discrete shipments were weak for Q2 YoY.
• Q2 of 2016 has overall shipments down 3.1% compared to the previous year. AMD saw growth in desktop discrete shipments shipments while Nvidia declined in desktop discrete shipments. Desktop discrete shipments were strong for Q2 YoY due to the weak Q2 2015 shipments.

There isn't anything strange or contradictory about AMDs statements AFAICS. Without knowing the actual shipped numbers that Mercury are using you can't make any statements using just marketshare alone. And that's even assuming Mercury has a good idea of how many chips AMD has shipped to AIBs (board partners) and OEMs (mostly notebook partners). But regardless, AMD are using Mercury's shipped numbers and not their market share numbers. You'd need to actually buy Mercury's report to know the actual numbers that AMD are using.

Regards,
SB

 

[Razor1]

The comparison is because that is what AMD did, they didn't compare directly to Q2 of 2015, they stated the numbers of GPU's shipped is larger than anything they have done since Q4 of 2014.

And yeah you do need to purchase the numbers to see the exact figures

We can see they grew marketshare, yes, by shipping more then last quarter, yes but also by nV losing ground by the lose of their own sales by competition from internally (pascal launch).

I can't verify that statement about most cards shipped/sold since q4 of 2014 unless I mix and match JPR and Mercury numbers.